Enigma of Karl Pearson and Bayesian Inference
Total Page:16
File Type:pdf, Size:1020Kb
Load more
Recommended publications
-
F:\RSS\Me\Society's Mathemarica
School of Social Sciences Economics Division University of Southampton Southampton SO17 1BJ, UK Discussion Papers in Economics and Econometrics Mathematics in the Statistical Society 1883-1933 John Aldrich No. 0919 This paper is available on our website http://www.southampton.ac.uk/socsci/economics/research/papers ISSN 0966-4246 Mathematics in the Statistical Society 1883-1933* John Aldrich Economics Division School of Social Sciences University of Southampton Southampton SO17 1BJ UK e-mail: [email protected] Abstract This paper considers the place of mathematical methods based on probability in the work of the London (later Royal) Statistical Society in the half-century 1883-1933. The end-points are chosen because mathematical work started to appear regularly in 1883 and 1933 saw the formation of the Industrial and Agricultural Research Section– to promote these particular applications was to encourage mathematical methods. In the period three movements are distinguished, associated with major figures in the history of mathematical statistics–F. Y. Edgeworth, Karl Pearson and R. A. Fisher. The first two movements were based on the conviction that the use of mathematical methods could transform the way the Society did its traditional work in economic/social statistics while the third movement was associated with an enlargement in the scope of statistics. The study tries to synthesise research based on the Society’s archives with research on the wider history of statistics. Key names : Arthur Bowley, F. Y. Edgeworth, R. A. Fisher, Egon Pearson, Karl Pearson, Ernest Snow, John Wishart, G. Udny Yule. Keywords : History of Statistics, Royal Statistical Society, mathematical methods. -
Two Principles of Evidence and Their Implications for the Philosophy of Scientific Method
TWO PRINCIPLES OF EVIDENCE AND THEIR IMPLICATIONS FOR THE PHILOSOPHY OF SCIENTIFIC METHOD by Gregory Stephen Gandenberger BA, Philosophy, Washington University in St. Louis, 2009 MA, Statistics, University of Pittsburgh, 2014 Submitted to the Graduate Faculty of the Kenneth P. Dietrich School of Arts and Sciences in partial fulfillment of the requirements for the degree of Doctor of Philosophy University of Pittsburgh 2015 UNIVERSITY OF PITTSBURGH KENNETH P. DIETRICH SCHOOL OF ARTS AND SCIENCES This dissertation was presented by Gregory Stephen Gandenberger It was defended on April 14, 2015 and approved by Edouard Machery, Pittsburgh, Dietrich School of Arts and Sciences Satish Iyengar, Pittsburgh, Dietrich School of Arts and Sciences John Norton, Pittsburgh, Dietrich School of Arts and Sciences Teddy Seidenfeld, Carnegie Mellon University, Dietrich College of Humanities & Social Sciences James Woodward, Pittsburgh, Dietrich School of Arts and Sciences Dissertation Director: Edouard Machery, Pittsburgh, Dietrich School of Arts and Sciences ii Copyright © by Gregory Stephen Gandenberger 2015 iii TWO PRINCIPLES OF EVIDENCE AND THEIR IMPLICATIONS FOR THE PHILOSOPHY OF SCIENTIFIC METHOD Gregory Stephen Gandenberger, PhD University of Pittsburgh, 2015 The notion of evidence is of great importance, but there are substantial disagreements about how it should be understood. One major locus of disagreement is the Likelihood Principle, which says roughly that an observation supports a hypothesis to the extent that the hy- pothesis predicts it. The Likelihood Principle is supported by axiomatic arguments, but the frequentist methods that are most commonly used in science violate it. This dissertation advances debates about the Likelihood Principle, its near-corollary the Law of Likelihood, and related questions about statistical practice. -
National Academy Elects IMS Fellows Have You Voted Yet?
Volume 38 • Issue 5 IMS Bulletin June 2009 National Academy elects IMS Fellows CONTENTS The United States National Academy of Sciences has elected 72 new members and 1 National Academy elects 18 foreign associates from 15 countries in recognition of their distinguished and Raftery, Wong continuing achievements in original research. Among those elected are two IMS Adrian Raftery 2 Members’ News: Jianqing Fellows: , Blumstein-Jordan Professor of Statistics and Sociology, Center Fan; SIAM Fellows for Statistics and the Social Sciences, University of Washington, Seattle, and Wing Hung Wong, Professor 3 Laha Award recipients of Statistics and Professor of Health Research and Policy, 4 COPSS Fisher Lecturer: Department of Statistics, Stanford University, California. Noel Cressie The election was held April 28, during the business 5 Members’ Discoveries: session of the 146th annual meeting of the Academy. Nicolai Meinshausen Those elected bring the total number of active members 6 Medallion Lecture: Tony Cai to 2,150. Foreign associates are non-voting members of the Academy, with citizenship outside the United States. Meeting report: SSP Above: Adrian Raftery 7 This year’s election brings the total number of foreign 8 New IMS Fellows Below: Wing H. Wong associates to 404. The National Academy of Sciences is a private 10 Obituaries: Keith Worsley; I.J. Good organization of scientists and engineers dedicated to the furtherance of science and its use for general welfare. 12-3 JSM program highlights; It was established in 1863 by a congressional act of IMS sessions at JSM incorporation signed by Abraham Lincoln that calls on 14-5 JSM tours; More things to the Academy to act as an official adviser to the federal do in DC government, upon request, in any matter of science or 16 Accepting rejections technology. -
The Likelihood Principle
1 01/28/99 ãMarc Nerlove 1999 Chapter 1: The Likelihood Principle "What has now appeared is that the mathematical concept of probability is ... inadequate to express our mental confidence or diffidence in making ... inferences, and that the mathematical quantity which usually appears to be appropriate for measuring our order of preference among different possible populations does not in fact obey the laws of probability. To distinguish it from probability, I have used the term 'Likelihood' to designate this quantity; since both the words 'likelihood' and 'probability' are loosely used in common speech to cover both kinds of relationship." R. A. Fisher, Statistical Methods for Research Workers, 1925. "What we can find from a sample is the likelihood of any particular value of r [a parameter], if we define the likelihood as a quantity proportional to the probability that, from a particular population having that particular value of r, a sample having the observed value r [a statistic] should be obtained. So defined, probability and likelihood are quantities of an entirely different nature." R. A. Fisher, "On the 'Probable Error' of a Coefficient of Correlation Deduced from a Small Sample," Metron, 1:3-32, 1921. Introduction The likelihood principle as stated by Edwards (1972, p. 30) is that Within the framework of a statistical model, all the information which the data provide concerning the relative merits of two hypotheses is contained in the likelihood ratio of those hypotheses on the data. ...For a continuum of hypotheses, this principle -
Memorial to Sir Harold Jeffreys 1891-1989 JOHN A
Memorial to Sir Harold Jeffreys 1891-1989 JOHN A. HUDSON and ALAN G. SMITH University of Cambridge, Cambridge, England Harold Jeffreys was one of this century’s greatest applied mathematicians, using mathematics as a means of under standing the physical world. Principally he was a geo physicist, although statisticians may feel that his greatest contribution was to the theory of probability. However, his interest in the latter subject stemmed from his realization of the need for a clear statistical method of analysis of data— at that time, travel-time readings from seismological stations across the world. He also made contributions to astronomy, fluid dynamics, meteorology, botany, psychol ogy, and photography. Perhaps one can identify Jeffreys’s principal interests from three major books that he wrote. His mathematical skills are displayed in Methods of Mathematical Physics, which he wrote with his wife Bertha Swirles Jeffreys and which was first published in 1946 and went through three editions. His Theory o f Probability, published in 1939 and also running to three editions, espoused Bayesian statistics, which were very unfashionable at the time but which have been taken up since by others and shown to be extremely powerful for the analysis of data and, in particular, image enhancement. However, the book for which he is probably best known is The Earth, Its Origin, History and Physical Consti tution, a broad-ranging account based on observations analyzed with care, using mathematics as a tool. Jeffreys’s scientific method (now known as Inverse Theory) was a logical process, clearly stated in another of his books, Scientific Inference. -
A Parsimonious Tour of Bayesian Model Uncertainty
A Parsimonious Tour of Bayesian Model Uncertainty Pierre-Alexandre Mattei Université Côte d’Azur Inria, Maasai project-team Laboratoire J.A. Dieudonné, UMR CNRS 7351 e-mail: [email protected] Abstract: Modern statistical software and machine learning libraries are enabling semi-automated statistical inference. Within this context, it appears eas- ier and easier to try and fit many models to the data at hand, thereby reversing the Fisherian way of conducting science by collecting data after the scientific hypothesis (and hence the model) has been determined. The renewed goal of the statistician becomes to help the practitioner choose within such large and heterogeneous families of models, a task known as model selection. The Bayesian paradigm offers a systematized way of as- sessing this problem. This approach, launched by Harold Jeffreys in his 1935 book Theory of Probability, has witnessed a remarkable evolution in the last decades, that has brought about several new theoretical and methodological advances. Some of these recent developments are the focus of this survey, which tries to present a unifying perspective on work carried out by different communities. In particular, we focus on non-asymptotic out-of-sample performance of Bayesian model selection and averaging tech- niques, and draw connections with penalized maximum likelihood. We also describe recent extensions to wider classes of probabilistic frameworks in- cluding high-dimensional, unidentifiable, or likelihood-free models. Contents 1 Introduction: collecting data, fitting many models . .2 2 A brief history of Bayesian model uncertainty . .2 3 The foundations of Bayesian model uncertainty . .3 3.1 Handling model uncertainty with Bayes’s theorem . -
Orme) Wilberforce (Albert) Raymond Blackburn (Alexander Bell
Copyrights sought (Albert) Basil (Orme) Wilberforce (Albert) Raymond Blackburn (Alexander Bell) Filson Young (Alexander) Forbes Hendry (Alexander) Frederick Whyte (Alfred Hubert) Roy Fedden (Alfred) Alistair Cooke (Alfred) Guy Garrod (Alfred) James Hawkey (Archibald) Berkeley Milne (Archibald) David Stirling (Archibald) Havergal Downes-Shaw (Arthur) Berriedale Keith (Arthur) Beverley Baxter (Arthur) Cecil Tyrrell Beck (Arthur) Clive Morrison-Bell (Arthur) Hugh (Elsdale) Molson (Arthur) Mervyn Stockwood (Arthur) Paul Boissier, Harrow Heraldry Committee & Harrow School (Arthur) Trevor Dawson (Arwyn) Lynn Ungoed-Thomas (Basil Arthur) John Peto (Basil) Kingsley Martin (Basil) Kingsley Martin (Basil) Kingsley Martin & New Statesman (Borlasse Elward) Wyndham Childs (Cecil Frederick) Nevil Macready (Cecil George) Graham Hayman (Charles Edward) Howard Vincent (Charles Henry) Collins Baker (Charles) Alexander Harris (Charles) Cyril Clarke (Charles) Edgar Wood (Charles) Edward Troup (Charles) Frederick (Howard) Gough (Charles) Michael Duff (Charles) Philip Fothergill (Charles) Philip Fothergill, Liberal National Organisation, N-E Warwickshire Liberal Association & Rt Hon Charles Albert McCurdy (Charles) Vernon (Oldfield) Bartlett (Charles) Vernon (Oldfield) Bartlett & World Review of Reviews (Claude) Nigel (Byam) Davies (Claude) Nigel (Byam) Davies (Colin) Mark Patrick (Crwfurd) Wilfrid Griffin Eady (Cyril) Berkeley Ormerod (Cyril) Desmond Keeling (Cyril) George Toogood (Cyril) Kenneth Bird (David) Euan Wallace (Davies) Evan Bedford (Denis Duncan) -
School of Social Sciences Economics Division University of Southampton Southampton SO17 1BJ, UK
School of Social Sciences Economics Division University of Southampton Southampton SO17 1BJ, UK Discussion Papers in Economics and Econometrics Professor A L Bowley’s Theory of the Representative Method John Aldrich No. 0801 This paper is available on our website http://www.socsci.soton.ac.uk/economics/Research/Discussion_Papers ISSN 0966-4246 Key names: Arthur L. Bowley, F. Y. Edgeworth, , R. A. Fisher, Adolph Jensen, J. M. Keynes, Jerzy Neyman, Karl Pearson, G. U. Yule. Keywords: History of Statistics, Sampling theory, Bayesian inference. Professor A. L. Bowley’s Theory of the Representative Method * John Aldrich Economics Division School of Social Sciences University of Southampton Southampton SO17 1BJ UK e-mail: [email protected] Abstract Arthur. L. Bowley (1869-1957) first advocated the use of surveys–the “representative method”–in 1906 and started to conduct surveys of economic and social conditions in 1912. Bowley’s 1926 memorandum for the International Statistical Institute on the “Measurement of the precision attained in sampling” was the first large-scale theoretical treatment of sample surveys as he conducted them. This paper examines Bowley’s arguments in the context of the statistical inference theory of the time. The great influence on Bowley’s conception of statistical inference was F. Y. Edgeworth but by 1926 R. A. Fisher was on the scene and was attacking Bayesian methods and promoting a replacement of his own. Bowley defended his Bayesian method against Fisher and against Jerzy Neyman when the latter put forward his concept of a confidence interval and applied it to the representative method. * Based on a talk given at the Sample Surveys and Bayesian Statistics Conference, Southampton, August 2008. -
The Enigma of Karl Pearson and Bayesian Inference
The Enigma of Karl Pearson and Bayesian Inference John Aldrich 1 Introduction “An enigmatic position in the history of the theory of probability is occupied by Karl Pearson” wrote Harold Jeffreys (1939, p. 313). The enigma was that Pearson based his philosophy of science on Bayesian principles but violated them when he used the method of moments and probability-values in statistics. It is not uncommon to see a divorce of practice from principles but Pearson also used Bayesian methods in some of his statistical work. The more one looks at his writings the more puzzling they seem. In 1939 Jeffreys was probably alone in regretting that Pearson had not been a bottom to top Bayesian. Bayesian methods had been in retreat in English statistics since Fisher began attacking them in the 1920s and only Jeffreys ever looked for a synthesis of logic, statistical methods and everything in between. In the 1890s, when Pearson started, everything was looser: statisticians used inverse arguments based on Bayesian principles and direct arguments based on sampling theory and a philosopher-physicist-statistician did not have to tie everything together. Pearson did less than his early guide Edgeworth or his students Yule and Gosset (Student) to tie up inverse and direct arguments and I will be look- ing to their work for clues to his position and for points of comparison. Of Pearson’s first course on the theory of statistics Yule (1938, p. 199) wrote, “A straightforward, organized, logically developed course could hardly then [in 1894] exist when the very elements of the subject were being developed.” But Pearson never produced such a course or published an exposition as integrated as Yule’s Introduction, not to mention the nonpareil Probability of Jeffreys. -
Frequentist and Bayesian Statistics
Faculty of Life Sciences Frequentist and Bayesian statistics Claus Ekstrøm E-mail: [email protected] Outline 1 Frequentists and Bayesians • What is a probability? • Interpretation of results / inference 2 Comparisons 3 Markov chain Monte Carlo Slide 2— PhD (Aug 23rd 2011) — Frequentist and Bayesian statistics What is a probability? Two schools in statistics: frequentists and Bayesians. Slide 3— PhD (Aug 23rd 2011) — Frequentist and Bayesian statistics Frequentist school School of Jerzy Neyman, Egon Pearson and Ronald Fischer. Slide 4— PhD (Aug 23rd 2011) — Frequentist and Bayesian statistics Bayesian school “School” of Thomas Bayes P(D|H) · P(H) P(H|D)= P(D|H) · P(H)dH Slide 5— PhD (Aug 23rd 2011) — Frequentist and Bayesian statistics Frequentists Frequentists talk about probabilities in relation to experiments with a random component. Relative frequency of an event, A,isdefinedas number of outcomes consistent with A P(A)= number of experiments The probability of event A is the limiting relative frequency. Relative frequency 0.0 0.2 0.4 0.6 0.8 1.0 020406080100 n Slide 6— PhD (Aug 23rd 2011) — Frequentist and Bayesian statistics Frequentists — 2 The definition restricts the things we can add probabilities to: What is the probability of there being life on Mars 100 billion years ago? We assume that there is an unknown but fixed underlying parameter, θ, for a population (i.e., the mean height on Danish men). Random variation (environmental factors, measurement errors, ...) means that each observation does not result in the true value. Slide 7— PhD (Aug 23rd 2011) — Frequentist and Bayesian statistics The meta-experiment idea Frequentists think of meta-experiments and consider the current dataset as a single realization from all possible datasets. -
Mathematical Statistics and the ISI (1885-1939)
Mathematical Statistics and the ISI (1885-1939) Stephen M. Stigler University of Chicago, Department of Statistics 5734 University Avenue Chicago IL 60637, USA [email protected] 1. Introduction From its inception in 1885, the International Statistical Institute took all of statistics as its province. But the understood meaning of “statistics” has undergone significant change over the life of the Institute, and it is of interest to inquire how that change, whether evolutionary or revolutionary, has been reflected in the business of the Institute, in its meetings and its membership. I propose to consider this question in regard to one particular dimension, that of mathematical statistics, and one particular period, that from the rebirth of international statistics in 1885 until the eve of World War II. Mathematical statistics has been a recognized division of statistics at least since the 1860s, when Quetelet at the Congress in the Hague admonished the assembled statisticians not to divide the école mathématique from the école historique, but to take statistics whole, to open the doors to statistics in the broadest sense in all its flavors (Stigler, 2000). But even with this recognition, mathematical statistics only emerged as a discipline later. I have argued elsewhere that this emergence can be dated to the year 1933, and even if one thinks that date absurdly precise, there was a distinct and palpable change that was evident around that time (Stigler, 1999, Ch. 8). It was then that changes initiated a half-century earlier by Galton and Edgeworth and developed subsequently by Karl Pearson and Udny Yule, were seized upon and transformed by Ronald Fisher and Jerzy Neyman and Egon S. -
Statistical Inference a Work in Progress
Ronald Christensen Department of Mathematics and Statistics University of New Mexico Copyright c 2019 Statistical Inference A Work in Progress Springer v Seymour and Wes. Preface “But to us, probability is the very guide of life.” Joseph Butler (1736). The Analogy of Religion, Natural and Revealed, to the Constitution and Course of Nature, Introduction. https://www.loc.gov/resource/dcmsiabooks. analogyofreligio00butl_1/?sp=41 Normally, I wouldn’t put anything this incomplete on the internet but I wanted to make parts of it available to my Advanced Inference Class, and once it is up, you have lost control. Seymour Geisser was a mentor to Wes Johnson and me. He was Wes’s Ph.D. advisor. Near the end of his life, Seymour was finishing his 2005 book Modes of Parametric Statistical Inference and needed some help. Seymour asked Wes and Wes asked me. I had quite a few ideas for the book but then I discovered that Sey- mour hated anyone changing his prose. That was the end of my direct involvement. The first idea for this book was to revisit Seymour’s. (So far, that seems only to occur in Chapter 1.) Thinking about what Seymour was doing was the inspiration for me to think about what I had to say about statistical inference. And much of what I have to say is inspired by Seymour’s work as well as the work of my other professors at Min- nesota, notably Christopher Bingham, R. Dennis Cook, Somesh Das Gupta, Mor- ris L. Eaton, Stephen E. Fienberg, Narish Jain, F. Kinley Larntz, Frank B.